Safe Superintelligence and NVIDIA have reached a cooperation and will use the Vera Rubin architecture. Safe Superintelligence (S

Safe Superintelligence reaches cooperation with NVIDIA and will use Vera Rubin architecture. Safe Superintelligence (SSI), an AI startup founded by former OpenAI chief scientist Ilya Sutskever, announced today that it has reached a long-term strategic cooperation agreement with chip giant NVIDIA (NVIDIA). According to the agreement, SSI will focus on using NVIDIA's next-generation Vera Rubin computing platform to obtain computing power support that is an order of magnitude higher than before. As part of the collaboration, Nvidia will make a "significant investment" in SSI, and the two parties will also conduct joint research and development around the underlying technologies of Nvidia's existing and future computing platforms. It is worth noting that Nvidia stated that it decided to promote this in-depth cooperation after being granted rare access to SSI’s highly confidential research results. Regarding this cooperation, Suzkevi said: "The research results we have are worthy of large-scale expansion, and obtaining a powerful NVIDIA computing cluster will help us achieve this goal. We are very proud to work with Huang Renxun and the NVIDIA team, and firmly believe that the heavy focus on the Vera Rubin platform will take us to a new level." Before this switch to NVIDIA, SSI had chosen to cooperate with the Google cloud (Google) in April 2025. Cloud) and used Google’s customized TPU accelerator to develop cutting-edge models. Google was also one of SSI’s investors at the time. SSI has not yet stated whether the new Vera Rubin deployment will completely replace the existing Google TPU infrastructure or run in parallel with it. Since its establishment, SSI has always maintained a high sense of mystery. It has not launched any commercial products and has not publicly released specific academic research details. However, in the current AI industry, it has become a common trend for top laboratories to adopt multi-vendor computing infrastructure in order to cope with the unprecedented demand for computing power. For example, OpenAI uses infrastructure from Microsoft, CoreWeave, AWS, Oracle and Google cloud at the same time. The hardware level covers a variety of chips and self-developed hardware such as Nvidia, AMD, Cerebras, AWS Trainium and so on. via cnBeta.COM - Chinese industry information station (author: source: cnBeta.COM)